K Manoj Kumar (@kmanoj296) 's Twitter Profile
K Manoj Kumar

@kmanoj296

Hacking AI & automation @VegaStack. Documenting my dopamine hits and misses on X.

ID: 2645546016

linkhttps://vegastack.com calendar_today14-07-2014 14:19:35

1,1K Tweet

271 Followers

224 Following

K Manoj Kumar (@kmanoj296) 's Twitter Profile Photo

Or go one step further - close all server ports including 22 for ssh Cloudflare Tunnel runs a daemon (cloudflared) on your server that makes outbound-only connections to Cloudflare. No inbound ports needed to be open at all. Traffic flows through Cloudflare, then through the

K Manoj Kumar (@kmanoj296) 's Twitter Profile Photo

Having worked with many businesses on automation over the last few years, the pattern I keep seeing is different from what those industry job posting % suggest Terraform at 88% sounds right but most teams hiring for it aren't struggling with HCL syntax. They're struggling with

K Manoj Kumar (@kmanoj296) 's Twitter Profile Photo

Been using Opus 4.6 for the past few hours now. It is like that dev who ships fast but you always review their PRs twice. #1 on filesystem, #3 on skills, same price as 4.5 but less token efficient. Got to manually tweak the effort levels each time. It's optimized for the win

Been using Opus 4.6 for the past few hours now. It is like that dev who ships fast but you always review their PRs twice. 

#1 on filesystem, #3 on skills, same price as 4.5 but less token efficient. Got to manually tweak the effort levels each time.

It's optimized for the win
K Manoj Kumar (@kmanoj296) 's Twitter Profile Photo

Human engineers know when to stop. AI agents don't. They optimize to finish the task, not for time. Exit rules > more prompting. P.S we automated our QA pipeline using CC skills, about 70% coverage on first run

Human engineers know when to stop. AI agents don't. 

They optimize to finish the task, not for time.

Exit rules > more prompting.

P.S we automated our QA pipeline using CC skills, about 70% coverage on first run
K Manoj Kumar (@kmanoj296) 's Twitter Profile Photo

Stopped using Opus 4.7 for coding 2w ago. Here's what I run now: GPT-5.5 for coding. Opus for reviews - mostly useless now, doesn't pick much, seems nerfed!? What changed my mind: - Codex is faster, I can't even notice context compaction. Its token efficient too. - It